Barrett’s oesophagus (BO) is an acquired premaligant condition that is the only known precursor to oesophageal adenocarcinoma (OA); a malignancy whose incidence has risen steadily in the Western world in the last 20–30 years.¹ The lifetime risk of developing an adenocarcinoma in the context of BO is 2–5%,² and despite recent advances in its treatment the associated morbidity and mortality remain high.³ BO advances through a series of morphological stages: from metaplasia through dysplasia of different grades and finally to adenocarcinoma—the metaplasia-dysplasia-adenocarcinoma sequence (MCS).^4,5 Although this genetic process is not as well outlined as in the colon,⁴ it is thought to be a similar multistep process consisting of genetic and epigenetic mutations, involving the same or a similar group of genes, which over many years leads to increasing genomic instability and ultimately to the formation of an autonomous clone of cells with invasive and metastatic properties.⁵

Knowledge of the natural history of the adenoma:carcinoma sequence in the colon has translated to clinical medicine, prompting the removal of adenomatous polyps at the premalignant stage, with an associated reduction in the incidence of colorectal cancer.^6,7 Regrettably, thus far, such gains have not been made in dysplasia in the context of BO, with the effectiveness of current surveillance strategies yet to be conclusively proven; a matter soon to be addressed by the BOSS (Barrett’s Oesophagus Surveillance Study), a study integrated with the AspECT (Aspirin and Esomprazole Chemoprevention Trial of Cancer in Barrett’s Oesophagus) (7500 patients in total).⁸ At present, a patient with BO can only be risk stratified on the basis of either endoscopically visible ulceration or by grade of dysplasia (negative, indefinite, low grade, high grade, or cancer) within the endoscopic biopsies.⁹ Neither of these methods is satisfactory: the inter and intra pathologist interpretation of dysplasia has proven hard to reproduce, with one study demonstrating only a 48% agreement in the grade of dysplasia in BO between expert pathologists; this, however, was increased to 86% when high grade dysplasia was compared with all of the preceding three grades.¹⁰ In today’s global health service, specialist pathologists are a scarce commodity, and a second opinion in all cases of dysplasia in BO stretches available resources, necessitating other methods by which patients with BO can be risk stratified.

In BO, the right point of intervention in the disease process is unclear, with the intervention itself associated with a significant morbidity and mortality. The current surveillance programme of endoscopy and biopsies is extremely labour intensive, and not without suffering to the patient. What is needed, therefore, is an indicator capable of identifying those patients at greatest risk of neoplastic progression and consequently requiring closer monitoring and earlier surgical intervention—a “biomarker”. Ideally, this tool would be sensitive and specific to BO, identifiable in the premalignant phase, and increase with malignant progression; it would be as non-invasive as possible (preferably detectable in blood, cytology, urine, or faeces) and it would be inexpensive. It could be used as a stand alone measurement but it could interface with existing techniques such as narrow band imaging or high magnification chromoendoscopy—its sole purpose may be to identify a cohort of patients who would benefit from the current screening programmes, with or without novel endoscopic techniques. As well as predicting neoplastic progression, it would also be beneficial if the biomarker identified those patients with dysplasia who would benefit from surgery, excluding those in whom surgery was too late. The urologists have prostate specific antigen, itself not without its limitations, and every gastroenterologist that has ever undertaken BO surveillance would now like a similar, or preferably better, tool.

In this issue of Gut, Murray and colleagues¹¹ describe the potential use of p53 immunostaining in oesophageal biopsies as such a biomarker (see page 1390). Accepting the unsatisfactory relationship between immunostaining for the protein and sequencing for mutations,¹² this study correlates with our knowledge of the molecular events in the progression from metaplasia to adenocarcinoma and presses the case for p53 one day being endorsed as a biomarker in BO. However, the findings of their paper indicate that for routine and pragmatic purposes this case is far from proven. Many other markers, both in BO and other areas of cancer, have been suggested but few if any have realised their initial promise: poor trial design, type of tissue selected, and technical processing of the material been shown to be biased in one way or another, or lacked reproducibility,¹³ have all played their part in discrediting potential candidates.¹⁴ Any such tools identified in this role need to be subjected to the highest quality control, and to this end the National Cancer Institute’s Early Detection Research Network (EDRN)¹⁵ outline a system for validation. It is comprised of five separate phases: from laboratory identification of promising markers in the preclinical phase 1, validation studies in phases 2 and 3, through the prospective clinical screening, sensitivity and specificity values in phase 4, and finally the impact on reduction of disease burden as part of a screening programme on the population in phase 5—the only true point at which clinical success can be quantified. Due to the natural history of neoplastic progression in the field of BO and the length of time taken to progress to adenocarcinoma, phase 5 studies would have to be large scale studies and to date no biomarker in BO has been evaluated at this level although the AspECT RCT is such a vehicle for this purpose. For this reason, in the field of BO, we feel it necessary to add phase 6 to this validation process; proven use in multiple centres and in multiple populations (table 1⇓). Only when a biomarker has met these standards could it be confidently used in daily clinical practice. Murray and colleagues¹¹ show p53 to have qualified at best to phase 3 or 4, but no further!

The vast majority of phase 1 and 2 studies result from application of high throughput transcriptional profiling and proteomics applied to the molecular knowledge of neoplastic progression in BO. The list of candidates is vast but with numerous mutations taking place in the setting of genomic instability the most challenging task is filtering out the “hitchhikers”—those markers which are just bystanders and do not impact on subsequent clonal expansion. Of the molecular events that take place in the MCS, those that have showed most potential as biomarkers, in addition to alterations in p53, are DNA ploidy and proliferation. P53 (17p) and P16 (9p) are both tumour suppressor genes and contribute to G1/S phase control of the cell cycle. The former regulates damage checkpoints in the cell cycle and therefore prevents cells with damaged DNA from dividing, in addition to maintaining chromosomal order at the time of mitosis. At a laboratory based level, previous studies have shown promise, with phase 1 and 2 studies demonstrating that the number of p53 lesions increases as the MCS sequence progresses,^16,17 and at a clinical level, a single phase 4 study¹⁸ has shown that 17p loss of heterozygosity (LOH) is a good predictor of progression, which together place p53 as one of the biomarkers with the most potential in the this field (table 2⇓). Within BO, a p16 mutation is one of the earliest mutations to take place, enabling clonal expansion to take place throughout the segment (table 2⇓)^19,20: 85–90% of BO contain one or more p16 lesions (a mutation, LOH, or methylation of the promoter)²⁰ but although it has been evaluated at phases 1 and 2, its role needs to be validated clinically in phases 3–5.

Proliferative markers, such as Ki-67, should ostensibly be good indices for progression but good data from phase 3, 4, or 5 translational studies are lacking (table 2⇑). The evidence base is greater for the potential use of cyclin D1, a promoter of cell division, and hence integral in the hyperproliferative state; levels have been shown to increase in a prospective phase 3, indicating risk of adenocarcinoma progression,²¹ as well as being overexpressed in phase 1 and 2 studies in OAs within BO²² but this has not been widely reproduced (table 2⇑). Another biomarker that increases with the progression from metaplasia to dysplasia is DNA ploidy abnormalities (aneuploidy and tetraploidy), as a result of increased genomic instability (table 2⇑). This can be detected with flow cytometry and has been shown in the context of BO, including in phase 4 studies, to be predictive of high grade dysplasia and adenocarcinoma,^18,23 Most likely because of technical difficulties, these results have also not been reproduced in other studies,²⁴ and as a result flow cytometry has not progressed to phase 5 trials.

The holy grail is not yet round the corner. Until then, we will continue using dysplasia, accepting that phase 4 studies of high grade dysplasia have shown it to be imprecise for identifying those likely to progress to carcinoma, with some demonstrating that most of these patients die of unrelated causes.²⁵ P53 shows some promise but clinicians will require large, prospective, preferably multicentre trials before its use is integrated to surveillance strategies in BO. Only 5% of patients who present with OA already have a diagnosis of Barrett’s metaplasia²⁶; the perfect biomarker should therefore be more accessible than through endoscopic biopsies—interestingly, antibodies to our prime candidate p53 can be detected in the serum of patients with OA.²⁷

The list of candidates being generated from phase translational 1 and 2 trials continues at an exponential rate, yet at present, as clinicians, we are left only with histopathology as a way to risk stratify our patients with BO. Translational science has failed to take these promising markers and deliver a robust tool to determine which patients we should be focusing our attentions on and which we should leave well alone. Why have none made the grade? What has been lacking so far is a strategic plan to identify the most promising biomarkers, and take them, and only them, through the different phases of validation on a large scale in a coordinated fashion. For this to occur, a well defined infrastructure of directed and optimised research is needed to allow the important scientific discoveries to be translated into large, population based, integrated, multicentre clinical trials as efficiently as possible.²⁸ Both the AspECT and BOSS trials offer this infrastructure. Half full? At least the level appears to be rising!